Improvement on the wideband DOA estimation by averaging covariance matrices over frequency bins

An improvement on the wideband DOA estimation is proposed. The new covariance matrix of a frequency bin is obtained by averaging the covariance matrices of its adjacent frequency bins, then the general wideband DOA algorithms are applied based on the new covariance matrices. Compared with the wideband algorithm without covariance matrices average, the improvement has the higher resolution and the smaller root mean square error for most cases. The theoretical analysis of the errors of DOAs shows that the average operation is effective and the error from this operation is negligible especially when very limited snapshots are available. The computer simulation is applied to compare the performance between the improved method and the conventional method.